Automatic clustering of media items (e.g., digital photographs, video) is useful for organizing and finding items in a library of many media items, whether a local collection of a single user's media or a global collection of many users' media. Currently, there are a number of algorithmic solutions using time differences alone to cluster items. But clustering media files by time does provide adequate clustering by events.